AI Coding for Web Developers
Use AI to write code faster — but only if you know how to steer it, verify output, and avoid the traps models fall into every day.
Last reviewed: June 2026
Tool features change quickly. Pages in this section include review dates — check them before relying on pricing, shortcuts, or UI labels.
Start Here
Common AI Coding Mistakes
Hallucinated APIs, phantom imports, and other mistakes models make, with fixes.
Cursor
Agent mode, @-mentions, rules, and when to use Plan vs Ask.
Prompting for Code
Structured prompts that produce usable diffs, not essays.
Building MCP Servers
Connect external tools and data to your AI coding environment.
Learning Paths
Solo developer: Day 1 with AI coding
- Common AI Coding Mistakes — know what goes wrong before you trust output
- Choosing an AI Coding Tool — pick Cursor, Codex, Claude Code, Copilot, or others
- Prompting for Code — spec → plan → implement → test
- Project Rules — stop repeating your stack in every prompt
- Verifying AI Output — pre-merge checklist
Optional depth: How LLMs Work → Context Engineering → Building MCP Servers
Team lead: Rolling out AI tools
- AI Coding Policy for Teams — approved tools, data handling, PR requirements
- Project Rules — shared AGENTS.md and rules rollout
- Responsible Use — licensing, bias, and user-facing norms
- Verifying AI Output — required gates before merge
- Security and Prompt Injection — when AI touches production features
Optional depth: Choosing a Tool (procurement section) → Security Anti-patterns
Start Here If…
| Your situation | Start with |
|---|---|
| AI suggested code that won't build | Common Mistakes |
| Picking your first AI coding tool | Choosing a Tool |
| Model keeps missing project conventions | Project Rules |
| Shipping a chat feature in your app | Streaming Chat Tutorial |
| Connecting a database to your agent | Building MCP Servers |
| Setting team policy from scratch | Team AI Policy |
| Large repo; model can't find the right file | RAG for Codebases |
Page Map
| Topic | Full guide | Quick reference |
|---|---|---|
| Mistakes and fixes | Common Mistakes | AI Coding Mistakes |
| Pre-merge review | Verifying Output | Pre-Merge Verification |
| Cursor | Cursor | Cursor Shortcuts |
| OpenAI Codex | OpenAI Codex | OpenAI API Cheat Sheet |
| Gemini / Google | Gemini Code Assist | Model Picker |
| AWS coding | Amazon Q Developer | — |
| Multi-agent IDE | Devin Desktop | — |
| Prompting | Prompting for Code | AI Prompting |
| MCP servers | Building MCP Servers | MCP Config |
| Agent skills | Agent Skills | AGENTS.md Template |
| Rules comparison | Rules vs Skills | — |
| Structured JSON | Structured Outputs | OpenAI API Cheat Sheet |
| PR review | AI Code Review | Pre-Merge Verification |
| Claude Code CLI | Claude Code | Claude Code Commands |
| Open-source agents | Cline, Aider | — |
| LLM API routes | LLM APIs | LLM API Route Handler |
| OpenAI API | OpenAI API | OpenAI API Cheat Sheet |
| Model selection | Choosing a Tool | Model Picker |
| Streaming chat app | Streaming Chat Tutorial | LLM API Route Handler |
| Team policy | Team AI Policy | — |
| Token budgets | Tokens and Context | — |
What This Section Covers
This is not a generic "what is machine learning" course. Web Reference AI focuses on what working developers need today:
- Tools — Cursor, Claude Code, Copilot, Codex, Gemini, Devin Desktop, Amazon Q, and how to pick the right one
- Prompting and context — Rules files,
@references, token budgets, and project conventions - Building with AI — MCP servers, LLM APIs, agentic workflows, and RAG for codebases
- Shipping AI features — Streaming, security, cost, and prompt injection in web apps
- Verification — Treating AI output as a draft, not a finished product
- Foundations — Mental models for tokens, context, and evaluating output
- Tutorials — End-to-end walkthroughs you can run locally
New to how models work under the hood? Start with How LLMs Work — a condensed primer, not a textbook chapter.
How These Pages Are Maintained
Pages are hand-authored Markdoc, reviewed on a rolling basis. Time-sensitive topics (tool UI, pricing, model IDs) include Last reviewed callouts. Prefer official vendor documentation linked from each page over stale inline model names. Section hubs list learning paths; pillar pages include runnable examples and primary-source links.
Cheat Sheets
Quick references you can keep open while coding: